Machine Learning to Production

I've completed a masters in CS focusing on machine learning. I've learned lots about algorithms at a theoretical level and implementing them in Java/Weka and in Python/Scikit-learn.

What I didn't learn and am slightly stumped on is moving ML to production.

That is to say I can get good results on data and someone wants to use my application to make daily predictions. How do you move ML applications to production so users can feed in their new input and receive predictions preferably with minimal input from myself.

(I understand this is a big and vague question - but it's a new area for me).